Weekly Tech+Bio Highlights #60: Eli Lilly & NVIDIA to Build AI Supercomputer for Drug Discovery
Anthropic brings Claude to Life Sciences, new NVIDIA-backed model may beat AlphaFold3, MIT ships open-source peptide binder & OpenFold gets a new preview
Hi! This is BiopharmaTrend’s weekly newsletter, Where Tech Meets Bio, where we explore technologies, breakthroughs, and cutting-edge companies.
If this newsletter is in your inbox, it’s because you subscribed, or someone thought you might enjoy it. In either case, you can subscribe directly by clicking this button:
🤖 AI x Bio
(AI applications in drug discovery, biotech, and healthcare)
🔹 Lilly unveils the world’s largest AI factory for drug discovery, designed to train large-scale biomedical foundation and frontier models; select models will be available via TuneLab, an AI platform offering biotech access to $1B worth of proprietary data. The factory also integrates physical and agentic AI to accelerate biomanufacturing, robotics, and in silico drug design.
🔹 Open-source alternative to AlphaFold expands to drug and RNA targets — The OpenFold Consortium has released a preview of OpenFold3, a foundation model for predicting 3D structures of proteins, nucleic acids, and drug-like molecules, trained on over 13 million synthetic and 300K experimental structures, offering full commercial access and modular design for rapid integration across biotech R&D.
🔹 Claude expands into life sciences R&D — Anthropic launched “Claude for Life Sciences,” upgrading its Claude Sonnet 4.5 model with domain-specific tools, scientific connectors, and custom skills to support tasks from research to regulatory workflows.
🔹 Tahoe releases 3B-parameter open-source model for cell biology — Tahoe-x1, a compute-efficient foundation model trained on 100M perturbed single cells, achieving SoTA performance in gene essentiality and cancer program prediction, with full open-source release to support translational discovery.
🔹 New AI model challenges AlphaFold 3 in drug discovery — Genesis Molecular AI (formerly Genesis Therapeutics) unveiled Pearl, a generative model for drug-protein structure prediction that reportedly surpasses AlphaFold 3 on internal and external benchmarks by leveraging synthetic physics-based data and NVIDIA-backed infrastructure for industrial deployment.
🔹 Toward Universal Binder Design — MIT expands Boltz AI suite with BoltzGen for universal therapeutic design. Building on Boltz-1 (structure prediction) and Boltz-2 (binding affinity), BoltzGen enables open-source, all-atom generation of protein and peptide binders across modalities and targets, with nanomolar hits experimentally validated by 26 partners.
🔬 Universal model for molecular design — A global academic team has released ODesign, a foundation model capable of designing proteins, nucleic acids, small molecules, and metal ions for any biological target, achieving 50x higher throughput than previous tools and compressing days of design work into hours.
🔹 AI kills colon cancer stem cells by reprogramming them — UCSD researchers used a machine learning framework to restore lost gene expression in tumor stem cells, triggering self-destruction and reducing recurrence risk in patient-derived models.
🔹 AI uncovers new ALS drug target — insitro, led by Daphne Koller, has advanced its first AI-discovered therapeutic target for ALS into small-molecule development in partnership with Bristol Myers Squibb, using human cell models and machine learning to identify and optimize disease-relevant interventions.
🔹 Cellarity shares Science blueprint for AI + transcriptomics drug discovery — The Flagship-founded startup published a framework in Science describing its AI-driven, single-cell transcriptomic platform to design cell state–correcting drugs. The approach improved hit recovery 13–17x over conventional screens. The team also open-sourced 1.26M-cell datasets to accelerate external benchmarking. First clinical candidate (for sickle cell) is in phase 1.
🔹 Genomics launches Mystra to scale human genetics in drug discovery — Unveiled at ASHG 2025, the AI-driven platform integrates genotype-phenotype data to improve target validation and trial design, aiming to boost R&D success rates and reduce drug development costs.
🔹 Modeling the “social networks” of cells — Bo Wang, Head of Biomedical AI at $1B AI drug discovery startup Xaira Therapeutics, and researchers from University Health Network and University of Toronto, have developed a graph-based deep learning tool that predicts cell-to-cell communication using single-cell RNA data and protein interaction networks, offering improved performance across diverse biological contexts.
🔹 Cell BioEngines joins NVIDIA for Startups to build AI platform for HLA matching — As first-gen cell therapies face commercial collapse, Cell BioEngines is betting on HLA-guided, off-the-shelf allogeneic models.
🔹 Understanding may be out of reach—but control isn’t — Kim Branson, Global Head of AI at GSK, argues that biology’s complexity may defy full reverse-engineering due to computational irreducibility and chaos, suggesting the future lies in controlling rather than fully understanding biological systems using AI-driven, experiment-in-the-loop approaches.
🔹 In GEN interview, David Baker separates AI protein design hype from reality — The Nobel laureate says designing proteins from scratch is now possible, but AI’s true impact on medicine depends on unraveling biology’s complexity.
🔹 Ion channels built from scratch — Scientists in David Baker’s lab at the Institute for Protein Design, led by Yulai Liu, have designed functional calcium-selective ion channels entirely from scratch using AI tools, demonstrating that even complex, poorly understood biological functions can now be engineered from first principles.
🔹 BioRender’s vetted visual tools now live inside Anthropic’s Claude for Life Sciences, enabling real-time generation of publication-ready figures and experimental diagrams.
This newsletter reaches over 10K industry professionals from leading organizations across the globe. To reserve your sponsor slot in one of the upcoming issues, contact us at info@biopharmatrend.com
💰 Money Flows
(Funding rounds, IPOs, and M&A for startups and smaller companies)
🔹 Novartis acquires RNA biotech Avidity for $12B to expand its neuromuscular RNA pipeline, paying a 45% premium in a major bet on RNA delivery technology and neuroscience.
🔹 UK-based Chemify raises $50M+ to scale AI-driven molecule manufacturing — Backed by Wing VC and Insight Partners, Chemify will expand its automated “Chemifarm” facilities and “Chemputation” platform for on-demand molecular design.
🔹 London-based GHO Capital raised over $2.9B for its fourth fund, its largest yet, to expand investments in scalable manufacturing, medtech, and health tech, with a growing focus on AI-enabled platforms and biopharma services.
🔹 At the start of this month, the NIH launched a major push to standardize organoids — With a $87M investment, the new SOM Center aims to make lab-grown human tissue models reproducible and scalable using AI, robotics, and open protocols, accelerating drug testing and reducing reliance on animal models.
⚙️ Other Tech
(Innovations across quantum computing, BCIs, gene editing, and more)
🔹 Consumer BCI hits research-grade standards — Neurable becomes the first consumer brain-computer interface to meet iMotions’ scientific validation criteria, marking a major leap toward accessible, high-fidelity neurotech after 14 years of R&D and military-backed validation.
🔹 Altman’s rumored Merge Labs eyes noninvasive brain-AI interface — OpenAI CEO Sam Altman is reportedly backing Merge Labs, a stealth venture developing ultrasound-based brain sensing to enable “think-to-ChatGPT” communication, aiming for a soft, read-only human-AI merge without surgery, contrasting Neuralink’s invasive approach. Read more on the topic in our BCI deep dive.
🔹 Kidney organoids take a leap toward physiological function — new study in Cell Stem Cell (Huang et al.) reports the most advanced kidney organoids yet. When implanted in mice, they integrated with host vasculature and began filtering blood.
🔹 From digital design to printed biomaterials — new open-source 3D printer dubbed BEAVER enables low-cost, dual-extrusion printing of biotic materials like pectin, chitosan, and cellulose.
🏛️ Bioeconomy & Society
(News on centers, regulatory updates, and broader biotech ecosystem developments)
🔹 Healthcare leads AI adoption, with startups capturing 85% of spend — Menlo Ventures reports U.S. healthcare AI use has surged 7–10x since 2023, now outpacing all other sectors at 2.2x the broader economy. Hospitals drive $1B of $1.4B in spend, targeting documentation, coding, and prior auth. Startups dominate, while EHR incumbents scramble to catch up.
🔹 China’s biotech velocity — Chinese team developed a novel circular RNA modality embedding aptamers for carrier-free delivery, then jumped from rodent studies to a 9-person FIH trial—collecting ~500k human single-cell transcriptomes to map immune responses.
🔹 Formulation as a design problem — Christine Allen, founder of Intrepid Labs, argues that AI and automation are transforming drug formulation from trial-and-error into a multi-objective design process, enabling faster, scalable, and patient-tailored therapies. Intrepid emerged from stealth in May 2025 with $11M.
🔹 10x Genomics, Roche, and Prognosys sue Illumina over spatial and single-cell patents — filing two federal lawsuits alleging Illumina infringed nine patents tied to spatial biology and single-cell tech, reigniting IP battles just months after settling with other rivals.
🚀 A New Kid on the Block
(Emerging startups with a focus on technology)
🔹 Flagship unveils $50M-backed AI startup targeting undruggable proteins — Expedition Medicines launches with a generative covalent chemistry platform combining quantum modeling and chemoproteomics to design small molecules for hard-to-drug targets, starting with a prostate cancer program under Flagship’s partnership with Pfizer.